Acoustic detection of a lubrication level of the components of a timepiece movement

The use of multi-mode piezo-acoustic detection and ultrasonic stethoscope techniques allows for precise lubrication detection and adjustment in watch components, addressing the inefficiencies of existing methods and enhancing assembly compliance and maintenance efficiency.

EP4459385A1Pending Publication Date: 2024-11-06RICHEMONT INTERNATIONAL SA
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
EP2023170982
Authority / Receiving Office
EP · EP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-05-02
Publication Date
2024-11-06

AI Technical Summary

Technical Problem

Current methods for lubrication detection in watch components lack precision in determining the quantity of lubricant and do not provide a suitable lubrication process, leading to inefficiencies and increased costs in manufacturing and after-sales services.

Method used

A diagnostic method using multi-mode piezo-acoustic detection and ultrasonic stethoscope techniques to identify the acoustic signature of watch mechanisms, allowing for accurate determination of lubricant quantity and quality, with the option to adjust lubrication levels through instructions for use.

Benefits of technology

This method enables precise detection and adjustment of lubrication, reducing effort and cost by ensuring optimal lubrication levels, minimizing wear, and preventing issues related to overfilling or underfilling, thereby improving assembly compliance and reducing maintenance needs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure IMGAF001_ABST
    Figure IMGAF001_ABST
Patent Text Reader

Abstract

Diagnostic method for detecting the amount of lubrication of a clockwork mechanism, comprising the following steps: - to carry out an acoustic detection of the mechanism in operation, - to obtain an acoustic signature (1, 2, 3) of the mechanism, - to determine, based on the acoustic signature (1, 2, 3) of the mechanism, the amount of lubrication of the mechanism.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] The present invention relates generally to the field of watchmaking and jewelry, more particularly to the field of lubrication detection on components and assemblies.

[0002] The aim of the present invention is to propose a new way of detecting the correct lubrication of watch components (or the quantity of lubricant) and to propose a diagnosis or a method of lubrication of these in order to make it compliant.

[0003] It is known in the prior art to use acoustic detection. For example, CH716023 discloses a condition evaluation program with an accuracy evaluation process, wherein the accuracy evaluation function is a function of accumulating the time during which the sound emitted by the watch escapement (ticking) is collected to calculate a cumulative ticking sound generation time, and to determine whether or not the cumulative ticking sound generation time exceeds a predetermined threshold value, and wherein the notification data output function is a function generating notification data recommending an overhaul of the watch when it is determined that the cumulative ticking sound generation time exceeds the predetermined threshold value during the accuracy evaluation process.

[0004] However, the prior art document relates to a revision notification based on cumulative operating time, which is unsuitable for monitoring during manufacturing or assembly or after-sales service. Furthermore, the prior art document does not disclose detection of lubrication of watch components, nor the quantity, nor does it propose a lubrication method or a lubrication instruction method.

[0005] Thus, it is desired to be able to offer a diagnosis allowing the quality and quantity of lubrication of watch components to be diagnosed and to propose an adjustment or corrective action in order to obtain a compliant assembly of components, while reducing the effort and cost necessary for this purpose for the watchmaker.

[0006] Multi-mode piezo-acoustic detection or even more advantageously acoustic and ultrasonic stethoscopic detection make it possible to achieve sufficient fidelity in accuracy and precision for such a diagnosis, as well as its ease of implementation.

[0007] Therefore, the invention relates, in a first aspect, to a diagnostic method for detecting a quantity of lubricant in a watch or jewelry mechanism, comprising the following steps: carry out an acoustic detection of the mechanism in operation, using an acoustic sensor, identify an acoustic signature of the mechanism in operation, based on an acoustic quantity extracted from the acoustic detection, compare the acoustic signature with a reference acoustic quantity, define or determine, based on the comparison of the acoustic signature, the quantity of lubricant in the mechanism.

[0008] In other words, the invention relates to a diagnostic method for detecting a quantity of lubricant of a clockwork mechanism, comprising the following steps: carry out an acoustic detection of the clockwork mechanism in operation, determine an acoustic signature of the clockwork mechanism in operation, determine, based on the acoustic signature of the clockwork mechanism, the quantity of lubricant in the mechanism.

[0009] The term acoustic signature is used to describe a combination of acoustic emissions. In addition, mechanisms and their components, in normal or degraded operation, can be described as having their own acoustic signatures or characteristic sound attributes, which can be used to study their condition, behavior, and physical location. The signature is the set of elements that allow, for example, a sensor to characterize the presence, type, or identity of an object, from a signal of the object collected by the sensor. Signature is thus understood to be the set of characteristic and recognizable features extractable from the signal of a sensor and allowing something to be attributed or identified. This signature therefore depends not only on the object to be characterized but also on the characteristics of the sensor and the way in which it is applied.For example, an acoustic sensor presents its sensitivity, its own noise, its time and frequency response, that is to say a set of more or less distorting or reducing characteristics. The more sensitive and linear the acoustic sensor is, its time and frequency resolution extended to the entire emission of the object, and the more stable it is in these qualities, the more the capture of a signature is faithful in accuracy and precision, that is to say clear and reproducible, and therefore carries reliable diagnostic information.

[0010] Particularly near-field acoustic and ultrasonic stethoscopic detection, which has very high qualities of bandwidth, spectral flatness, low noise and immunity to parasitic disturbances, makes it possible to propose a diagnostic method for detecting or characterizing the quantity of lubricant, in particular at the time of manufacture and assembly, but also at any time, so as to know more precisely the condition of the watch and to be able to take corrective actions if necessary, such as for example lubricating the components more, removing excess lubricant or adding an additive in particular in order to limit the effect of this excess, planning maintenance on a scheduled date, predicting degraded operation and from when, without being limited to these examples. In other words, it is a question of detecting a posteriori the quantity of lubricant after having lubricated the mechanism.

[0011] Thus, it is possible to propose a process for diagnosing the quality and quantity of lubricant in watch components and to propose an adjustment or corrective action in order to obtain a compliant assembly of components, while reducing the effort and cost required for this purpose for the watchmaker.

[0012] In addition, a large number of returns to after-sales service are linked to wear phenomena which may be linked to a lack or poor arrangement of lubricants initially (i.e. at the time of manufacture) on certain watch mechanisms, in particular for self-winding mechanisms.

[0013] Reliability is generally greater the larger the quantity of lubricant, as long as an overflow does not risk causing the lubrication to leak and contaminate third-party surfaces or cause excessive running-in times. In other words, the quantity of lubricant should be adjusted, i.e. neither too much nor too little, and in the right place (for example between two parts in relative motion with respect to each other). Lubrication control can take place immediately after the lubrication step, by optical or visual inspection, before other assembly elements are assembled and cover the area to be checked. Lubrication can take time to establish itself, as capillary phenomena and the operation of the mechanism are driving parameters for this establishment. The final lubrication control after start-up and running-in is difficult to carry out because few contacts remain visible.Any disassembly to check compliance will tend to change the position of the lubrication with a risk of soiling surfaces and involves significant effort in terms of time and labor, and also risks damaging the parts. The check can be facilitated by the use of fluorescent lubrications to detect their presence in the area to be lubricated, but without giving any indication of the correct arrangement in the contact. The lubrication may be in an inappropriate place (for example above, below or next to it). Therefore, the method of the present invention has the advantage of solving these problems.

[0014] Advantageously, the method comprises the following step: carry out acoustic detection of sufficient characterization fidelity in accuracy and precision of the clockwork mechanism in operation, using an acoustic sensor.

[0015] Advantageously, the method comprises the following step: extract a combination of acoustic magnitude from the acoustic signature.

[0016] Advantageously, the method comprises the following step: extract spectral flatness from the acoustic signature.

[0017] This allows for effective and efficient detection of watch component lubrication, high detection accuracy and finesse, while having good reproducibility and limiting false positives or false negatives. Spectral flatness is one quantity among other acoustic quantities. This example of spectral flatness is based on a model with this parameter; other examples may not be based on it.

[0018] Advantageously, the mechanism is a mechanical watch movement.

[0019] Advantageously, the mechanism comprises a plurality of components, the method comprising the following step: defining or determining an amount of lubricant on the at least one of the plurality of components.

[0020] Advantageously, the mechanism comprises the plurality of components, and the method comprises the following step: defining or determining the amount of lubricant on each of the plurality of components.

[0021] This makes it possible to diagnose which component is operating normally or degraded, and to take corrective actions on this single component (or set of components) so as to limit the correction effort.

[0022] Advantageously, the method comprises the following step: defining or determining the amount of lubricant on at least one component based on a predetermined operating speed of the at least one component.

[0023] Advantageously, it is possible to provide a predetermined type of oil depending on the predetermined operating speed of the at least one component.

[0024] This allows the determination to be refined based on a relevant operating parameter.

[0025] Advantageously, the mechanism comprises at least one component chosen from a barrel, a winding mechanism, an escapement, an anti-shock system, cogs, an eccentric pivot, another pivot, a wolf tooth set arranged to cooperate with a pawl, the method comprising the following step: define or determine a quantity of lubricant on the at least one component.

[0026] This makes it possible to diagnose which component is operating normally or degraded, and to take corrective actions on this single component (or set of components) so as to limit the correction effort.

[0027] Advantageously, the amount of lubricant is determined from insufficient lubrication, adequate lubrication or excess lubrication.

[0028] Advantageously, the quantity of lubricant is a lubrication volume, preferably from 0.5 to 100 nL.

[0029] This allows you to detect good lubrication, especially for wolf teeth or an escape wheel.

[0030] Advantageously, the mechanism comprises a plurality of lubricating oils, the method comprising the following step: defining or determining the amount of lubricant of at least one lubricating oil among the plurality of lubricating oils.

[0031] Advantageously, the oils are for example the Moebius references: 9504 (synthetic grease for high friction, with a viscosity of 305 cSt at 20 degrees), 9501 (synthetic grease for high precision micromechanics with a viscosity of 176 cST at 20 degrees), 8200 (natural grease for watchmaking, for example for slow mobiles), Synth HP1300 (precision synthetic oil for high pressure, with a viscosity of 1250 cST at 20 degrees), Synth HP 500 (precision synthetic oil for high pressure, with a viscosity of 500 cST at 20 degrees), and / or 9010 available commercially, without being limited to these examples.

[0032] Advantageously, the method comprises the following step: defining or determining the amount of lubricant in each of the plurality of lubricating oils.

[0033] This allows for a more refined understanding of the mechanism, in order to determine a lubrication level, a quantity of lubricant and which oil is causing difficulties.

[0034] Advantageously, the method comprises the following step: use a confusion matrix to determine, based on the acoustic signature of the mechanism, the quantity of lubricant in the mechanism.

[0035] A confusion matrix is ​​a matrix that includes prediction results. It compares the actual data for a target variable with those predicted by a model. Correct and incorrect predictions are revealed and divided into classes, allowing them to be compared with defined values. In other words, in supervised machine learning, the confusion matrix is ​​a matrix that measures the quality of a classification system.

[0036] This makes the diagnosis more robust, improves the accuracy of detection, better classifies components and their lubrication and, in particular, limits false negatives and false positives.

[0037] A second aspect of the invention relates to a lubrication method for lubricating a watch or jewelry mechanism comprising the following steps: carry out or implement the diagnostic method according to the first aspect, issue a usage instruction or an adjustment instruction to a user.

[0038] Advantageously, the operating instruction may be to remove or add lubricant.

[0039] This makes it possible to offer a lubrication process with instructions for use or adjustment instructions to enable a user to correctly lubricate watch components.

[0040] Thus, it is possible to propose a process for correctly lubricating watch components, in particular in terms of the quality and quantity of lubricant for the watch components, and to propose an adjustment or corrective action in order to obtain a compliant assembly of the components, while reducing the effort and cost required for this purpose for the watchmaker.

[0041] Advantageously, the operating or adjustment instruction may be to add lubricant (lubricating oil or other), to remove lubricant, in particular if it appears that the movement is over-lubricated, to change the lubricant, to add an additive in order to modify the lubrication, to specify in what quantities these additions or removals should be made, to remove the component in question, for example to clean it entirely or to replace it.

[0042] Advantageously, the usage or adjustment instructions are displayed on a connected device, such as a connected magnifying glass.

[0043] This makes it possible to offer operation with a connected device, thus making the watchmaker's work easier.

[0044] This makes it possible to provide a lubrication (and lubrication adjustment) process and a human-machine interface to a user, in particular a watchmaker, in order to reduce his mental and physical effort for adjusting the assembly, while making it possible to improve the quality, precision and repeatability of the assembly. A connected magnifying glass is known for example from document CH717263. Such an electronic watchmaker's magnifying glass is intended to observe a watchmaking product, in particular a watch mechanism, and is equipped with: a magnifying lens, for observing the watch product, communication means, for receiving data, a display screen, for displaying the data, a beam splitter, designed to simultaneously direct towards a user's eye at least one image of the watch product coming from the magnifying lens and at least one image from the display screen. The electronic magnifier may further comprise an acquisition camera, for acquiring at least one image of the watch product, the communication means being designed to send to a remote user the image of the watch product acquired by the acquisition camera, the beam splitter being designed to simultaneously direct towards the acquisition camera at least one image of the watch product coming from the magnifying lens and at least one image from the display screen.

[0045] A third aspect concerns the use of an acoustic sensor to detect the quantity of lubricant in a watch or jewelry mechanism.

[0046] This involves using a sensor to detect the quantity and quality of lubrication.

[0047] Advantageously, the method further comprises the following step: using a sensor, preferably a piezo-acoustic sensor, to capture a plurality of sounds from the clockwork mechanism in operation and obtain an acoustic signature of the mechanism.

[0048] This allows sounds to be captured in order to implement the diagnostic process. The piezo-acoustic sensor has a good signal-to-noise ratio over a wide frequency band. It is also possible to provide multi-mode operation with this piezo-acoustic sensor in order to obtain a high signal-to-noise ratio over a wide frequency range via the natural vibration modes. This has the advantage of being able to go to high frequencies, improve the speed of rise of the sound signal and obtain better temporal precision. Measurement via direct contact (structure-borne transmission, for example by a crown) is possible. It is also possible to cut off low frequencies in order to isolate ambient noise.

[0049] Advantageously, the method comprises the following step: using a sensor, preferably a sensor with a plurality of ultrasonic microphones, for example digital, in order to capture the plurality of sounds and obtain an acoustic signature of the mechanism.

[0050] MEMS (microelectromechanical system) technology allows for a wide bandwidth (e.g. 100Hz - 80kHz) to be covered, while allowing for near-field contactless measurement (stethoscope type). The plurality of sensors may include four sensors, for example.

[0051] For chronometry-type measurements, it is possible to cut the low frequencies in order to isolate ambient noise.

[0052] It is possible to install the watch on a structure comprising the sensor with one (or more) acoustic access(s), preferably a felt, a coupling microcavity and one (or more) acoustic access(s) in order to allow the sensor to have greater sensitivity to the sounds of the mechanism.

[0053] The various sensors mentioned above allow for finer, more faithful listening, and therefore greater recognition capacity by obtaining a greater quantity of information, by extending the band towards high frequencies which contain useful acoustic energy, by obtaining information less disturbed by the acoustic (and vibratory) environment, by the fact of the rapid dissipation of short wavelengths in air and in solids, by obtaining information less disturbed by the connection and electrical environment, by digitizing the signals as close as possible to the transduction source, by obtaining very low self-noise by multiplying paired transducers, by the possibility of implementing coherent reconstruction, time shift, convolution, filtering and focusing techniques,that is to say more precisely all the methods of signal processing and acoustic antenna algorithms by microphone arrays and applicable in near field. In addition, by the use of miniature transducers of known sensitivity and factory matched in amplitude and phase, and for their entire lifetime, the technology does not require initial or periodic calibration.

[0054] Advantageously, the method further comprises the following step: use an acquisition device to record the captured sound data.

[0055] Advantageously, the method further comprises the following step: use a computer in particular to compare the captured sound data.

[0056] This allows the process to be implemented.

[0057] The sensors may be used in all aspects of the present invention.

[0058] Advantageously, the method further comprises the following step: submit the acoustic signature to a prediction machine to determine the amount of lubricant in the clockwork mechanism.

[0059] Advantageously, the prediction machine comprises a processor and / or a computer.

[0060] Advantageously, the prediction machine further comprises prediction software, such as artificial intelligence.

[0061] Advantageously, the prediction machine comprises one or more neural networks, such as one or more convolutional neural networks.

[0062] Advantageously, the prediction machine can determine the amount of lubricant and categorize the mechanism as compliant or non-compliant, based on the acoustic signature. Thus, the prediction machine must solve the problem of the amount of lubrication on the parts of the mechanism (including all lubrication variants described here).

[0063] The prediction machine then receives the acoustic signature, uses the model, and determines the quantity of lubricant.

[0064] This learning phase allows the construction of calibrated reference data for later comparison during a prediction / production phase with defect detection. In particular, learning allows the acquisition of test data from reference parts or parts tested / simulated in parallel to build the model. The reference parts can be confirmed by a watchmaker.

[0065] The prediction machine uses a predictive model that can be built from repeated and trained operation.

[0066] Thus, it is possible to train the model for detection and then use the model, with or without model improvement.

[0067] Advantageously, the reference quantity is predetermined.

[0068] Advantageously, the reference quantity is defined by the model.

[0069] To establish a prediction model that can receive as input the acoustic signature and give as output a judgment or determination of a compliant or non-compliant mechanism and an adjustment or use instruction. It is possible to provide, during the learning phase, the provision of data relating to the actual sound signals tested, including compliant and non-compliant mechanisms. The model may include a correlation phase in which the acoustic signature is linked to the conclusion of the compliant or non-compliant mechanism.

[0070] Once the learning phase is complete, we can move on to a prediction phase, for example during a mechanism control process.

[0071] A fourth aspect of the present invention relates to a method of training a prediction machine, comprising the following steps: operating a clockwork or jewelry mechanism, recording a plurality of sounds emitted during operation of the clockwork mechanism, training the prediction machine using a model and a series of sounds emitted by the mechanism in operation, so as to determine an amount of lubricant, preferably on one or more of a plurality of components of the mechanism.

[0072] Advantageously, the model is defined by the construction of the mechanism.

[0073] Advantageously, the model is predetermined by construction of the mechanism.

[0074] Advantageously, the model is adapted after operating the mechanism or a plurality of mechanisms, so as to modify parameters or coefficients of the constructed model.

[0075] Advantageously, the learning process is carried out by machine learning.

[0076] Advantageously, the learning method includes the use of neural networks.

[0077] The input data is the caliber of the movement, the constituent parts, the geometry of the parts, the type of oil, and / or the quantity of oil expected.

[0078] The output data is the quantity of oil measured, the qualification as insufficient, correct or excessive lubrication, and / or the determination of the part on which the lubrication is inspected.

[0079] This makes it possible to propose a learning process in order to improve the determination of the quantity of lubricant in the clock movement.

[0080] Other characteristics and advantages of the present invention will appear more clearly on reading the detailed description which follows, of embodiments of the invention given by way of non-limiting example and illustrated by the appended drawings, in which: therefigure 1 represents an acoustic detection of a clock mechanism in operation, the figure 2 represents a spectral flatness of the clock mechanism in operation, the figure 3 represents an example of a measuring device for measuring a sound and implementing the diagnostic method, the figure 4 represents another example of a measuring device for measuring sound and implementing the diagnostic method.

[0081] There figure 1 represents an acoustic detection of a watch mechanism (or movement or caliber) in operation. The graph of the figure 1 represents the evolution of the sound level 10 (on the ordinate, expressed in decibels dB) as a function of the sound frequency (on the abscissa, expressed in Hertz Hz).

[0082] The graph represents three curves: a first curve 1 represents the evolution of the noise level 10 for lubrication at zero percent of the reference value, that is to say that curve 1 represents the case without lubrication. A second curve 2 represents the evolution of the noise level 10 for lubrication at forty percent of the reference value, that is to say that curve 2 represents the case of ideal or target lubrication. A third curve 3 represents the evolution of the noise level 10 for lubrication at one hundred percent of the reference value, that is to say that curve 3 represents the case of excess lubrication.In other words, the first curve 1 is an acoustic signature for lubrication at zero percent of the reference value, the second curve 2 is an acoustic signature for lubrication at forty percent of the test reference value, and the third curve 3 is an acoustic signature for lubrication at one hundred percent of the reference value. Note that the reference value is predetermined, and can also be called the test reference value, control reference value, monitoring value, alert value, or non-conformity value. Note that the value of forty percent can be adapted and chosen, for example, between zero and one hundred, such as twenty-five or fifty. Also note that it is generally more damaging to the mechanism to be without lubricant than with excess lubricant.

[0083] It is thus possible to determine the quantity of lubricant based on the acoustic signature of the mechanism.

[0084] In addition, it is possible to submit an acoustic signature of the mechanism to a prediction machine. The prediction machine then makes it possible to determine the quantity of lubricant in the mechanism.

[0085] The machine may use a model established by learning, if necessary initially predetermined in relation to the intrinsic construction of the mechanism, and / or adapted on the basis of the operation of the mechanism. The learning step may further rely on human confirmation of the quantity of lubricant to teach the model and the machine how to classify mechanisms and determine the quantity of lubricant. The learning step may confirm or highlight that for the given mechanism it is possible to use spectral flatness as a relevant acoustic quantity, as detailed with reference to the figure 2 .

[0086] There figure 2 represents a spectral flatness of the clock mechanism in operation.

[0087] The graph represents three candlesticks: a first candlestick 11 represents the spectral flatness of the sound level 10 for lubrication at zero percent of the reference value, that is to say that the candlestick 11 represents the case without lubrication of the first acoustic signature 1. A second candlestick 12 represents the spectral flatness of the sound level 10 for lubrication at forty percent of the reference value, that is to say that the candlestick 12 represents the case of ideal lubrication of the second acoustic signature 2. A third candlestick 13 represents the spectral flatness of the sound level 10 for lubrication at one hundred percent of the reference value, that is to say that the candlestick 13 represents the case of excess lubrication of the third acoustic signature 3.

[0088] Extreme values ​​can be discarded (shown here as dotted lines) to make detection more reliable.

[0089] The first candlestick 21 comprises a first median value 21, the second candlestick 22 comprises a second median value 22, the third candlestick 23 comprises a third median value 23.

[0090] The median values ​​21, 22, 23 allow to determine, with the help of acoustic detection, what is the lubrication of the mechanism, that is to say whether it is a mechanism not lubricated, partially lubricated or correctly lubricated, or excessively lubricated. It is possible to make this determination also from all the data of candlesticks 11, 12, 13 and not only from the median values.

[0091] Note that it is also possible to use an average or any other characteristic instead of the median, or in addition.

[0092] This makes it possible to offer an acoustic control (or acoustic diagnosis), preferably carried out during the final control operations (at the end of manufacturing), which makes it possible to report on the conformity of the lubrication of the watch mechanism. The watch mechanism is, for example, an automatic winding system.

[0093] It is possible to apply the diagnosis to any part of the mechanism, for example to the eccentric pivot.

[0094] An example of a method to qualify or calibrate the diagnosis is to provide three series of movements assembled according to three different quantities of lubrication: ten movements without lubrication, ten movements very well lubricated (at nominal), five or ten movements lubricated in an intermediate manner, for example between twenty-five percent and fifty percent.

[0095] The movements are then placed on a movement listening bench, which can be rotated in both directions (clockwise or counterclockwise). It is possible to use a listening bench without rotation.

[0096] The automatic system is then put into operation and the sound it emits is captured.

[0097] The operating signal is recorded, processed (for example by spectral decomposition, or by power or by energy of the signal or combination) and described by its temporal and / or frequency acoustic signature.

[0098] The objective is to define acoustic markers associated with the lubrication level and to define an average (or median or characteristic) signal describing the lubrication category (without lubrication, partially lubricated or correctly lubricated). It is also possible to obtain more details on the volume of lubricant, the type of oil, the component for which the lubrication is determined, as indicated previously. It is thus possible to train a model so as to teach it to correctly detect the quantity of lubricant. For example, it is possible to use a classification method of the "random forest" type to determine that spectral flatness is one of the most relevant characteristics of analysis, and to build or train a model, for example of the GLM type "Generalized linear model".

[0099] Each measured movement is then classified or placed in its category (for example poorly lubricated, partially lubricated, correctly lubricated, or even excess lubrication by including in the control batch movements with excess lubricant).

[0100] Spectral flatness can be seen as a measure of the amount of noise in a signal, or conversely, of the tonality. It is possible to use an acoustic quantity other than spectral flatness, such as loudness, frequency, duration, power, but not limited to these examples. It can be an acoustic quantity detected as relevant by artificial intelligence or machine learning.

[0101] If the spectral flatness is close to zero, we are dealing with a spectrum with more "tonal" peaks from an acoustic point of view.

[0102] If the value is higher, we are dealing with a rather flat spectrum, or closer to noise.

[0103] This makes it possible to classify or place movements in the correct category, and more specifically to ensure that no non-lubricated movement is placed in the correctly lubricated category and vice versa.

[0104] In more detail, spectral flatness or tone coefficient, can also be called Wiener entropy, and is a measure used in signal processing to characterize an audio spectrum.

[0105] Spectral flattening is usually measured in decibels and quantifies the extent to which a sound resembles a pure tone, as opposed to (or contrasts with) noise. In other words, tonality is the amount of peaks or resonant structure in a power spectrum, as opposed to the flat spectrum of white noise.

[0106] High spectral flatness (close to "one" for white noise) indicates that the spectrum has a similar amount of power in all spectral bands—this looks like white noise, and the spectrum graph would appear relatively flat and smooth. Low spectral flatness (close to zero for a pure tone) indicates that the spectral power is concentrated in a relatively small number of bands: this typically looks like a mixture of sine waves, and the spectrum appears "spiky."

[0107] Spectral flatness is defined by dividing the geometric mean of the power spectrum by the arithmetic mean of the power spectrum, i.e. Flatness = ∏ n = 0 N − 1 x n N ∑ n = 0 N − 1 x n N = exp 1 N ∑ n = 0 N − 1 ln x n 1 N ∑ n = 0 N − 1 x n where x(n) represents the magnitude of bin number n. A single (or more) empty bin gives a flatness of zero, so this measure is most useful when bins are not generally empty.

[0108] The ratio produced by this calculation is often converted to a decibel scale for ratios, with a maximum of zero dB and a minimum of -∞ dB. Spectral flatness can also be measured in a specific subband, rather than across the entire band.

[0109] It is further possible to provide an artificial intelligence learning step in order to determine a physical parameter of another part to be diagnosed and to apply the diagnostic method to this parameter in order to determine whether this other part is correctly lubricated, poorly lubricated, or partially lubricated and to apply the different variant embodiments of the invention to this part.

[0110] In particular, clusters should be determined using machine learning to detect the lubrication of this other part. Neural networks can also be used for this purpose.

[0111] There figure 3 represents an example of a measuring device for measuring a sound and implementing the diagnostic method.

[0112] As mentioned above, it is important when manufacturing mechanical watches to check the operation of the various components of the watch as assembled. This includes, for example, checking the correct synchronization of the clockwork mechanism (or clock movement), the presence of manufacturing defects, etc.

[0113] In general, inventors have recognized that the near-field acoustic emission of a watch can be highly variable along its outer casing and that a spatial scan of the near-field acoustic emission provides a plurality of information that allows for better characterization of the acoustic behavior and the noise and emission sources of the watch.

[0114] The inventors also recognized that a measurement based on one or more microphones and spatial scanning provides the best available information in terms of frequency band, because microphones have an inherently wider measurement bandwidth than vibration sensors, and also that a plurality of microphones provides the best available information from weak signal captures, because the combination of a plurality of signal captures inherently provides increased signal-to-noise ratio capability.

[0115] As illustrated in figure 3 , a measuring device 2000 for measuring the sound generated by the watch W may be used, such as the measuring device 2000 having: a support surface 1100, a plurality of microphones 1200 placed on or under the support surface 1100, a controller connected to the plurality of microphones 1200, wherein the controller is configured to process the output of the plurality of microphones 1200 so as to obtain one or more signals representative of the sound generated by the watch W.

[0116] The plurality of microphones 1200 can be placed in close proximity to the object to be measured, such as a watch W, so that environmental noise is limited from reaching the microphones 1200 and the sound of the watch is measured with very high sensitivity and selectivity.

[0117] In some embodiments of the measurement device 2000, the plurality of microphones 1200 may be phase and amplitude matched. The plurality of microphones 1200 may be configured to have an amplitude shift of less than 3 dB, preferably less than 2 dB, more preferably less than 1 dB, and / or the plurality of microphones 1200 may be configured to have a phase shift of less than 10°, preferably less than 5°, more preferably less than 1°.

[0118] The plurality of microphones 1200 may be MEMS-based microphones.

[0119] The plurality of microphones 1200 may be separated from each other by an intra-microphone distance, the intra-microphone distance being less than 1 / 2 of the smallest wavelength measured by the measuring device 2000, preferably less than 1 / 4 of the smallest wavelength, more preferably less than 1 / 8 of the smallest wavelength.

[0120] The plurality of microphones 1200 may cover a total distance, the total distance may be less than the smallest wavelength measured by the measuring apparatus 2000, preferably less than half of the smallest wavelength.

[0121] The plurality of microphones 1200 may cover a total distance, the total distance may be less than 10 cm, preferably less than 5 cm, more preferably less than 3 cm.

[0122] The plurality of microphones 1200 may comprise at least 4 microphones, preferably at least 16 microphones, more preferably at least 32 microphones.

[0123] Different microphones 1200, or different groups of microphones 1200, may be configured to measure sounds over different frequency ranges. In use, a volume V between the watch and a majority, preferably all, of the microphones may be filled with a liquid and / or gel.

[0124] The watch W to be measured or listened to may be placed at a measurement distance from the plurality of microphones 1200 which may be less than 1 cm, preferably less than 0.5 cm, more preferably less than 1 mm. The plurality of microphones 1200 may be configured to implement an acoustic antenna. An acoustic antenna may be considered as a device comprising an array of microphones 1200 used to detect sound waves coming from a particular direction.

[0125] Additionally, even though a microphone can measure in the bandwidth of 500 Hz to 200 kHz, it can be used to measure sounds in a narrower bandwidth only, for example, 500 Hz to 1 kHz, for optimal use.

[0126] The measuring device 2000 may be configured so that the watch W is held at the measuring distance without additional support. This may be achieved by providing a suitable frame for holding the watch W. Alternatively, or in addition, the support surface 1100 of the measuring device 2000 may be configured to hold the watch in place. The portion of the measuring device 2000 that holds the watch W in place may have a shape complementary to at least a portion of the outer shape of the watch, so that the watch may be held firmly and securely, while reducing the amount of ambient noise that reaches the microphone(s) due to the complementarity of the support structure and the shape of the watch, as will be illustrated with reference to the figure 4 .

[0127] The measuring device 2000 may also comprise a housing 2100 configured to define a volume V. The volume V is therefore an internal volume of the housing 2100. The housing 2100 may comprise or be connected to the support surface 1100, such that the plurality of microphones 1200 may be placed inside the volume V.

[0128] In this configuration, in order to allow sounds coming from the watch W, which is external to the volume V, to reach the microphones 1200, the support surface 1100 may comprise at least one opening 1110.

[0129] There figure 4 represents another example of a measuring device for measuring sound and implementing the diagnostic method.

[0130] As seen for example in the figure 4 , the support surface 4100 of the measuring device 2000 may be flexible and / or configured to conform to a surface of the watch W. The flexible support surface 4100 may be held in place by a suitable non-flexible support frame 4400, which is only schematically illustrated in figure 4 .

[0131] It will be understood that various modifications and / or improvements obvious to those skilled in the art may be made to the various embodiments of the invention described in the present description.

Claims

1. Diagnostic method for detecting a quantity of lubricant in a watch or jewelry mechanism, comprising the following steps: - carrying out an acoustic detection of the mechanism in operation, using an acoustic sensor, - identifying an acoustic signature (1, 2, 3) of the mechanism in operation, based on an acoustic quantity extracted from the acoustic detection, - comparing the acoustic signature (1, 2, 3) with a reference acoustic quantity, - defining, based on the comparison of the acoustic signature (1, 2, 3), the quantity of lubricant in the mechanism.

2. Diagnostic method according to the preceding claim, further comprising the following step: - extracting a combination of acoustic quantities (11, 12, 13) from the acoustic signature (1, 2, 3).

3. Diagnostic method according to the preceding claim, further comprising the following step: - extracting a spectral flatness (11, 12, 13) from the acoustic signature (1, 2, 3).

4. Diagnostic method according to one of the preceding claims, in which the mechanism comprises a plurality of components, the method comprising the following step: - determining a quantity of lubricant on at least one component of the plurality of components.

5. Diagnostic method according to the preceding claim, comprising the following step: - determining the quantity of lubricant on at least one component based on a predetermined operating speed of the at least one component.

6. Diagnostic method according to one of the preceding claims, in which the mechanism comprises at least one component chosen from a barrel, a winding mechanism, an escapement, an anti-shock system, a valve, cogs, an eccentric pivot, another pivot, a wolf tooth set arranged to cooperate with a pawl, the method comprising the following step: - determining a quantity of lubricant on the at least one component.

7. Diagnostic method according to one of the preceding claims, wherein the quantity of lubricant is determined from insufficient lubrication, adequate lubrication or excess lubrication.

8. Diagnostic method according to one of the preceding claims, in which the quantity of lubricant is a lubrication volume.

9. Diagnostic method according to one of the preceding claims, wherein the mechanism comprises a plurality of lubricating oils, the method comprising the following step: - determining the quantity of lubricant of at least one lubricating oil among the plurality of lubricating oils.

10. Diagnostic method according to the preceding claim, comprising the following step: - determining the quantity of lubricant of each oil of the plurality of lubricating oils.

11. Diagnostic method according to one of the preceding claims, comprising the following step: - using a confusion matrix in order to determine, as a function of the acoustic signature of the mechanism, the quantity of lubricant in the mechanism.

12. Lubrication method for lubricating a watch movement comprising the following steps: - carrying out the diagnostic method according to one of the preceding claims, - issuing a usage instruction to a user.

13. Lubrication method according to the preceding claim, in which the usage instruction is displayed on a connected device, such as a connected magnifying glass.

14. Method for training a prediction machine, comprising the following steps: - operating a watch or jewelry mechanism, - recording a plurality of sounds emitted during operation of the mechanism, - training the prediction machine using a model and a series of sounds emitted by the mechanism in operation, so as to determine a quantity of lubricant.

15. Method for learning a prediction machine according to the preceding claim, in which the model is predetermined by construction of the mechanism.

Citation Information

Patent Citations

  • A device for measuring the parameters of a watch, and a watch including such a device.

    CH708942A2

  • Electronic watchmaker's loupe, set of electronic watchmaker's loupes and method for carrying out an inspection of a watch product, in particular a watch mechanism.

    CH717263A1

  • Condition assessment program and watch bracelet to run the program.

    CH716023A2